-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 If a machine has very obvious memory corruption due to bad ram, is there anything beyond scrub that can verify the integrity of a pool? Am I correct in assuming that scrub will fix checksum errors, but not metadata errors? - -- Christopher Gorski mailto:cgorski at cgorski.org http://cgorski.org -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAkek2qIACgkQzUExD/R9ryrgPACeNGBt0Sw6VtF3lZfzO1apL283 QhUAni1n/XukShvJfv/SY4s8e8e/vbxi =+8il -----END PGP SIGNATURE-----
Hello Christopher,
Saturday, February 2, 2008, 9:03:30 PM, you wrote:
CG> -----BEGIN PGP SIGNED MESSAGE-----
CG> Hash: SHA1
CG> If a machine has very obvious memory corruption due to bad ram, is there
CG> anything beyond scrub that can verify the integrity of a pool? Am I
CG> correct in assuming that scrub will fix checksum errors, but not
CG> metadata errors?
Both metadata and data ara checksummed in zfs.
Then metadtaa are kept in 2 or three copies, depending on type.
Your data resilience depends on pool configuration obviously - so zfs
will always be able to detect data corruption and possibly fix it if
your pool has redundant config.
However if your memory is corrupting data then if data were corrupted
before application issued write then from zfs point of view everything
will be fine. If it was corrupted later then zfs should detect it and
correct in redundant config.
--
Best regards,
Robert Milkowski mailto:milek at task.gda.pl
http://milek.blogspot.com
>is there anything beyond scrub that can verify the integrity of a pool?Probably not. scrub will, at least, traverse all of the metadata involved in the tree of blocks. I don''t believe, however, that it checks integrity beyond that (e.g. of the space maps). If you know that your machine had bad RAM, your best bet is probably to use zfs send/restore (on a working system) to re-create as much metadata as possible. (If you need the data at all, given that it may well have undetected errors.) This message posted from opensolaris.org